An advanced gas discrimination method utilizing the periodic operation of a semiconductor gas sensor

This paper proposes a concentration-independent inflammable gas discrimination method utilizing transient response patterns of a semiconductor gas sensor. In the proposed method, the heater voltage of the sensor is changed periodically for a certain time interval after a gas sample is introduced to obtain accurate and uniform information from the response patterns regardless of the surrounding temperature variations. To realize concentration-independent gas discrimination, the DC components of the periodic response patterns which depend largely on the variation of ambient temperature and gas concentration are removed by applying a Fourier transform, and only AC components which show little change on their shape in spite of the same variations are used for pattern matching. The method is examined in discriminating five kinds of inflammable gases with a three layered backpropagation neural network, and average 99% discrimination rate is achieved